Add a physical detector interpretation to DetectorHeatMap and rethink the cost computation

[GilesStrong]

Currently it is unclear how the GMM learnt by DetectorHeatMap can be turned into physical detectors, this needs to be added. This should also feed into the cost calculation; the cost currently uses the mean of the sigmas of the GMM, rather than the sum (I understand this is for optimisation purposes, but it's unclear if the cost can then truly be measured in currency units). Additionally cost ignores the normalisation of the GMM PDF.